We should check if UMA is sent out when --enable-automation is in place. It should not be, so there should be nothing to do. If it is, we should not be sending uma out while automated. We should still collect it though because it is accessible over apis imwhile automating.

what does the flag do? Does it enable the devtools debugging protocol for input or somehting?

--enable-automation is enabled for telemetry, chromedriver, puppeteer and other automation scenarios. I'm not aware of the clients that would use input domain w/o using the --enable-automation. And if they do, they should not be distinguishable from the real users, that's the purpose of this api.

To understand that flag more I have a few more questions.

Can I have devtools debugging protocol on and send the input commands myself with some REST client without having --enable-automation flag for Chrome? Do the events get injected still?

How about another scenario which is running Chrome with --enable-gpu-benchmarking. This enabled the GPU benchmarking API to be accessed within a webpage. So the web page can start injecting inputs. Can I have this scenario without --enable-automation?

Pavel, do you have any opinion about the cases I mentioned?